fix(documents): keep an incomplete extraction on the document

A long document read in parts can end with a part cut at the provider's
output limit, a part never read, or more facts than one review takes.
That was told once, as a notice when the run finished; a review opened
later looked complete.

InboundDocument.extractionIncomplete (migration 0483) is written with
the facts by both stagings, the extract route or run and lab
auto-staging. The document detail publishes it and the review shows the
same note whenever it is opened. The disaster-recovery backup carries
it; an older file restores it as false.

Refs #1257

src/lib/openapi/routes/documents.ts

@@ -328,6 +328,10 @@ const inboundDocumentDetail = inboundDocument
     summary: z.string().nullable(),
     summaryGeneratedAt: z.string().nullable(),
     summaryState: z.enum(DOCUMENT_SUMMARY_STATES),
+    extractionIncomplete: z.boolean().meta({
+      description:
+        "Whether the staged facts may leave part of the document out: a part of a long document was cut at the provider's output limit or not read, the text or pages ran past what the reads take, or the document held more facts than one review takes. Kept from the last extraction (the extract route, its background run, or automatic lab staging) until the next one replaces the facts; false for a document never extracted. A review shows it whenever it is opened, not only when the run ends.",
+    }),
   })
   .meta({
     id: "InboundDocumentDetail",